PROCESSING OF MULTI-TEMPORAL SATELLITE DATA FOR LOCUST MASS
BREEDING CONTROL IN THE WEST SIBERIA
A. A. Tronin
Scientific Research Centre for Ecological Safety, RAS, Korpusnaya, 18, St-Peterburg, 197110, Russia tronin@at1895.spb.edu
Commission VI, WG VI/4
KEY WORDS: Multi-temporal satellite data, locust mass breeding, vegetation index
ABSTRACT:
The problem of locust mass breeding in Russia became more and more urgent. Last outbreaks of a locust in the West Siberia,
Kazakhstan and the Caucasus have evoked considerable interest of scientists to remote sensing methods of locust numbers control
and forecast. A few physical parameters determine the life cycle of the locust: temperature and humidity of the soil and air, green
biomass volume, soil type. Modern remote sensing has potential ability to retrieve these parameters. Some of these parameters,
mostly vegetation index, are successfully used for locust control, for example, in North Africa. Wide range of sensors can be used
for locust control. Long range of locust mass breeding (years) demands heritage satellite systems. At the moment only
NOAA(AVHRR) plus EOS(MODIS) meets these requirements. South part of the West Siberia was analysed for vegetation index
change in connection with locust spreading. Multi-temporal analysis was done on three levels: year, season and day to day. First was
multi-year range of observations. Vegetation index change from 1981 to 2005 was compiled. Years with vegetation index minimum
conform to maximum of locust numbers. Vegetation index changes in time like sinusoid with a period about 10 years.
Meteorological observations did not show any periodical processes for the first view. Long term air temperature analysis indicates
the growth at 4 degree in last 150 years. The sine shape of vegetation index looks like solar activity curve. Another type of analysis
was done in year cycle. Vegetation index change for 30 test sites in 2005 was calculated. Ten days MODIS composite images were
used for NDVI retrieving. Two field campaigns were done in 2005 for verification. Daily monitoring on the base of MODIS data
was applied to Madagascar test site, April 2002 locust mass breeding. Future multi-temporal data analysis will include surface
temperature, heat fluxes on the soil surface, soil humidity. For soil and DTM research ASTER, Landsat and SRTM data will be
included.
1. INTRODUCTION
The problem of locust mass breeding in Russia became more
and more urgent. Last outbreaks of a locust in the West Siberia,
Kazakhstan and the Caucasus have evoked considerable interest
of scientists to remote sensing methods of locust numbers
control and forecast. A few physical parameters determine the
life cycle of the locust: temperature and humidity of the soil and
air, green biomass volume, soil type. Modern remote sensing
has potential ability to retrieve these parameters. Some of these
parameters, mostly vegetation index, are successfully used for
locust control, for example, in North Africa, Australia
(Bryceson, 1990, Despland, Rosenberg and Simpson, 2004,
Hamilton and Bryceson, 1993, Tucker, Hielkema and Roffey,
1985).
The main aim of our research is application of satellite data for
monitoring and in future forecast of locust mass breeding in
West Siberia.
The south of West Siberia is one of the main agricultural areas
of Russia. The area has continental climate with cold winter and
dry hot summer. Year mean temperature is about 3.5 ºC, annual
precipitation is about 350 mm whereof about 100 mm is snow.
Steppe landscapes are widely spread. Most of the land is
plugged and the farming is based on grain-crops. Local test site
located on the border Russia/Kazakhstan area, in Novosibirsk
region. Test site landscapes are representative for wide
agricultural are in south part of West Siberia.
A few types of locust are founded in West Siberia: Italian locust
(Calliptamus italicus), Morocco locust (Dociostaurus
maroccanus) and Asian locust (Locusta migratoria migratoria
). The most dangerous is Italian locust. More than 4 million ha
was affected by locust in Kazakhstan and West Siberia in 1999.
Locust density was reached to 150 per square metre.
2003
2001
1999
1997
1995
1993
1991
1989
2003
2001
1999
1997
1995
1993
1991
1989
1987
Year
1985
Ground meteorological data were used in data analysis. Two
field campaigns were done for locust counting.
450 mm
400
350
300
250
200
150
100
50
0
1983
2.2 Ground data
Figure 3. Mean year temperature.
1981
MODIS and AVHRR satellite data were used for vegetation
index retrieving. Archive AVHRR monthly reflectance data for
band 1 and 2 were downloaded. AVHRR data have time range
since 1981 till 1998. MODIS/TERRA monthly reflectance data
from 1999 were also selected. 8-day composite of MODIS
bands 1 and 2 reflectance were involved in processing for 2005
year season. Daily MODIS bands 1 and 2 reflectance was used
for daily monitoring.
1987
2.1 Satellite data
1985
Year
1981
2. DATA
4.80 С
4.30
3.80
3.30
2.80
2.30
1.80
1.30
0.80
0.30
1983
Figure 1. Test site location. Zones: 1-taiga, 2-semi-taiga, 3steppe, 4-semi-desert, 5-test site.
Various types of satellite data were tested for locust mass
breeding monitoring: surface temperature, vegetation index, and
soil humidity. At the moment the vegetation index looks the
most promising parameter.
3. METHOD
Figure 4. Mean year precipitation – solid line, snow - dash line.
NDVI was retrieved on the base of reflectance in bands 1 and 2
both AVHRR and MODIS data (McGinnis, Tarpley, 1985).
Then mean values of NDVI were calculated for test site area.
Mean values were used for time-series analysis of NDVI data.
The only relation has been found – with sun spots numbers and
solar radiation (fig. 5).
Test site for year level monitoring located in relatively large
area: 53-55 N, 76-79 E. Year level monitoring indicates that
vegetation index changes in time like sinusoid with a period
about 10 years (fig. 2). Years with vegetation index minimum
2005
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
4.1 Year monitoring
Figure 5. Sun spots numbers.
4.2 Seasonal monitoring
Seasonal monitoring was executed for smaller size test site.
This test site has 30 test points. NDVI calculated for every 8
days during 2005 year vegetation season. Example for five test
points is represented on figure 6. Two field campaigns was
carried out
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
28.09.2005
14.09.2005
31.08.2005
17.08.2005
03.08.2005
20.07.2005
06.07.2005
22.06.2005
08.06.2005
25.05.2005
Year
11.05.2005
conform to maximum of locust numbers. Meteorological
observations did not show any periodical processes for the first
view (fig. 3, 4).
NDVI
27.04.2005
Figure 2. Vegetation index, monthly mean - August
1981-2000 – AVHRR, 2001-2005 – MODIS
Columns indicate periods of locust mass breeding.
1.00
0.90
13.04.2005
2005
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
1983
Year
30.03.2005
0.75 NDVI
0.7
0.65
0.6
0.55
0.5
0.45
0.4
0.35
0.3
1981
Year
1981
Results of NDVI analysis in relation with locust extension will
be presented on three levels: year level, seasonal level and daily
monitoring.
Sun spots
1983
4. RESULTS
180.3
160.3
140.3
120.3
100.3
80.3
60.3
40.3
20.3
0.3
Figure 6. Vegetation index change on local test points in 2005.
on test site in 2005 for locust counting. Probable correlation
was established for NDVI and locust density (fig. 7).
Locust per m2
12
10
8
6
4
2
NDVI
0
0.40
0.42
0.44
0.46
0.48
Figure 9. NDVI 25 April 2002 in Madagascar. Arrows indicate
locust outburst area.
Figure 7. Average vegetation index and locust density in 2005.
4.3 Daily monitoring
Due to the luck of locust last years on the test site, we tried to
test other areas affected by locust. One of the well recorded
outburst of locust took place in Madagascar in 2002 (Franc A.,
2006). Test site is shown on figure 8.
Figure 10. NDVI 26 April 2002 in Madagascar. Arrows indicate
locust outburst area.
5. DISCUSSION
First results of NDVI data application for locust mass breeding
look promising. Very interesting results were obtained on year
and daily level. Future researches will concentre on the relation
of NDVI with precipitation and air temperature. Land surface
temperature time-series analysis will be done. More complex
researches like surface heat fluxes analysis, surface thermal
inertia retrieving also will carry out. Soil humidity data also
require precise analysis.
6. CONCLUSION
Figure 8. Test site in Madagascar.
Daily vegetation index was retrieved for month period from 1
till 30 April 2002. Only two images from month time-series are
shown on figures 9 and 10. Vegetation index indicates
significant reduction from day to day. Ground truth data
confirm decrease of vegetation and locust mass breeding.
The results of multi-temporal satellite NDVI data on the year
level:
1) Vegetation index in the south part of the West Siberia is
changed periodically;
2) The period is about 11 year;
3) Locust mass breeding is coincided with minimum values of
vegetation index;
4) Remote sensing could be used for locust mass breeding
forecast on year level.
Seasonal level monitoring did not show clear results yet. Daily
level monitoring looks more promising:
1) Vegetation index indicates locust mass breading;
2) Rapid drop of the vegetation index points to locust mass
breading area;
3) Remote sensing could be used for locust mass breeding
monitoring in real time.
7. ACKNOWLEDGEMENTS
This research was executed in the framework of project #2908
funded by International Science and Technology Center.
Special thank to Alex Franc (CIRAD) for information and
photos from Madagascar.
REFERENCES
Bryceson, K. P., 1990. Locating locust infestation areas by
satellite. In: Proc. 5th Australian Remote Sensing Conference,
October 1990, pp. 581-584.
Despland, E., Rosenberg, J., and Simpson, S. J., 2004.
Landscape structure and locust swarming: a satellite' eye view.
Ecography, 27, pp. 381–391.
Hamilton, J. G. and Bryceson, K. P., 1993. Use of enhanced
GMS weather satellite data in locust forecasting. In: Pest
Control and Sustainable Agriculture, CSIRO, Melbourne, pp.
444-448.
McGinnis, D. F., Tarpley, J. D., 1985. Vegetation cover
mapping from NOAA/AVHRR. Advances in Space Research, 5
(6), pp. 359-369.
Franc A., 2006. Personal communication. CIRAD / Unite de
Recherche d'Acridologie Campus International de Baillarguet
(TA 40/D) 34398 Montpellier Cedex 5 - FRANCE
Tucker, C.J., Hielkema, J.U., and Roffey, J., 1985. The
potential of satellite remote sensing of ecological conditions for
desert locust survey and forecasting. International Journal of
Remote Sensing, 6, pp.127-138.